-
Notifications
You must be signed in to change notification settings - Fork 12.8k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Make TLS accesses explicit in MIR #71192
Conversation
Is there a good place where Miri-the-engine could assert that the pointers it encounters in constants do not point to thread-locals? Those Or even better, can we make it so that a thread-local static does not have a |
What I mean by this is that |
src/librustc_codegen_llvm/common.rs
Outdated
@@ -260,6 +261,7 @@ impl ConstMethods<'tcx> for CodegenCx<'ll, 'tcx> { | |||
Some(GlobalAlloc::Function(fn_instance)) => self.get_fn_addr(fn_instance), | |||
Some(GlobalAlloc::Static(def_id)) => { | |||
assert!(self.tcx.is_static(def_id)); | |||
assert!(!self.tcx.has_attr(def_id, sym::thread_local)); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You added this to a lot of places that fetch from the alloc_map
; any reason you didn't add it at the place where things get inserted?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
the insertion functions themselves don't have access to the tcx
. I could do it at the call sites, but since the only interesting call site is the one in librustc_mir_build, and that one already checks for thread locals and explicitly does not insert, an additional assertion would make no sense.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Ah, that's a shame, but makes sense. Maybe just add a comment there then?
I am happy this assertion holds at all. :D
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I could use thread locals in create_static_alloc
, but that seems odd and overkill.
Alternatively I can make the AllocMap
private to TyCtxt
and add functions to TyCtxt
that forward to the AllocMap
. That will cause a lot of churn, so I'd rather not do it in this PR.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
A comment in create_static_alloc
sounds enough.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I did the full thing in #71508
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
You mean, adding a comment? I cannot see it there, am I looking in the wrong place?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
No, I mean the forwarding of the functions. Once the other PR is merged, I can rebase this PR and add the assertion in create_static_alloc
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh I see.
Presumably, alloc_map.create_static_alloc
also still exists though, and its doc comment should ideally mention this condition as well.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It does not. There's only the TyCtxt
method, which forwards to reserve_and_set_dedup
, which is the place that will actually be doing the checking.
cc @nikomatsakis @pnkfelix @matthewjasper for the borrowck side of this (we used to have some hacks for thread-local |
The job Click to expand the log.
I'm a bot! I can only do what humans tell me to, so if this was not helpful or you have suggestions for improvements, please ping or otherwise contact |
☔ The latest upstream changes (presumably #71049) made this pull request unmergeable. Please resolve the merge conflicts. |
_ecx: &mut InterpCx<'mir, 'tcx, Self>, | ||
did: DefId, | ||
) -> InterpResult<'tcx, AllocId> { | ||
throw_unsup!(ThreadLocalStatic(did)) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I thought about errors in these default implementations again. Currently, we are inconsistent.
int_to_ptr
and abort
(and with this PR thread_local_alloc_id
) have default implementations that error "unsupported".
binary_ptr_op
, box_alloc
, ptr_to_int
also error in both const-eval and const-prop, but with separately implemented error messages.
I am not sure what's better, but it seems odd not to be consistent here.
@oli-obk I am wondering if you think an MCP would be appropriate for this PR? It seems to me like 'probably but not necessarily' -- a quick skim of the rustc-dev-guide, for example, suggests that it's just below the level of detail at which we document the MIR. That said, I think I at least would appreciate a few paragraphs of summary of what the high-level details of the approach are, and I think they'd probably make a nice addition to the rustc-dev-guide. Context: I've twice now clicked to view the diffs and I realize I'm spending most of the time just doing a "first pass" to try and extract the high-level view of 'what the heck is going on here'. I have some guesses but I don't know, and it'd be nicer if I didn't have to do that. |
I know FCP and RFC, what's MCP? |
major changes proposal: rust-lang/compiler-team#209 |
a11f92b
to
081ad6c
Compare
What is the status of this? |
Fix #[thread_local] statics as asm! sym operands The `asm!` RFC specifies that `#[thread_local]` statics may be used as `sym` operands for inline assembly. This also fixes a regression in the handling of `#[thread_local]` during monomorphization which caused link-time errors with multiple codegen units, most likely introduced by rust-lang#71192. r? @oli-obk
Fix link error with #[thread_local] introduced by rust-lang#71192 r? @oli-obk
Fix link error with #[thread_local] introduced by rust-lang#71192 r? @oli-obk
Fix link error with #[thread_local] introduced by rust-lang#71192 r? @oli-obk
Fix link error with #[thread_local] introduced by rust-lang#71192 r? @oli-obk
Fix link error with #[thread_local] introduced by rust-lang#71192 r? @oli-obk
Fix link error with #[thread_local] introduced by rust-lang#71192 r? @oli-obk
Fix link error with #[thread_local] introduced by rust-lang#71192 r? @oli-obk
Fix link error with #[thread_local] introduced by rust-lang#71192 r? @oli-obk
…ulacrum [beta] backports This PR backports the following: * Make novel structural match violations not a `bug` rust-lang#73446 * linker: Never pass `-no-pie` to non-gnu linkers rust-lang#73384 * Disable the `SimplifyArmIdentity` pass rust-lang#73262 * Allow inference regions when relating consts rust-lang#73225 * Fix link error with #[thread_local] introduced by rust-lang#71192 rust-lang#73065 * Ensure stack when building MIR for matches rust-lang#72941 * Don't create impl candidates when obligation contains errors rust-lang#73005 r? @ghost
r? @rust-lang/wg-mir-opt
cc @RalfJung @vakaras for miri thread locals
cc @bjorn3 for cranelift
fixes #70685